Dear Andrea, (copy to DSpace Developers List, DSpace General List to involve Repository Managers, and Mark Diggory)
I read your message to DSpace Developers List with attention. Please allow me some suggestions important for one of the next DSpace challenges: Subject (oriented) Repositories. In those repositories, knowledge discovery may be organised using a "flexible" matrix of concepts arranged along different facets (axis of the discovery matrix). Background information from IFLA: http://www.ifla.org/node/1297 Faceted browsing is very important in such a context. I would like to remind: * my presentation in Goteborg that I was honoured that you and Mark attended: ** DSUG 2009: http://gupea.ub.gu.se/handle/2077/21341 ** PDF: http://dsug09.ub.gu.se/index.php/dsug/dsug09/paper/viewFile/22/3 * an operational public DSpace site: http://www.windmusic.org * a Repository about Acute Exposure Toxicology (not public): http://www.poisoncentre.be * the project ASKOSI.org to create a SKOS framework and integrate it into DSpace (or other Java applications): ** Sub-projects description: http://www.destin.be/ASKOSI/Wiki.jsp?page=Projects ** General presentation: http://www.askosi.org/askosi_presentation.pdf Authority Lists are a functional subset of thesauri and classification systems. I believe DSpace (for Subject Repositories) must bring complete support for multilingual thesauri (their hierarchies, their synonyms, ...). Today, DSpace+ASKOSI: * imports PubMed records without loosing information (no "DC" oversimplification), * takes the SKOS version of the MeSH (http://thesauri.cs.vu.nl/eswc06/) or its database representation (UMLS) * controls the content of the keywords field, MeSH qualifiers included * allows retrieval by any synonyms or translations of a given concept * propagates indexation so retrieval can encompass all specific headings (like PubMed Search). * visualizes tables of the MeSH (and other configured thesauri) concepts together with the usage frequency in all managed applications (DSpace, Wikis, others: please look at ASKOSI presentation PDF, pages 8, 9 and 16): managing the (internal/external) vocabularies and their (local/remote) usage. Data abstraction proposed: * Concepts as SKOS entities implementing and benefiting from their properties (http://www.w3.org/2004/02/skos/) * Concept references represented by "words" (individual tokens) inserted in the text flow of DSpace fields. Important explanations: http://www.destin.be/ASKOSI/Wiki.jsp?page=Referring%20to%20Concepts Qualifiers can be put around concept references: * to precise their role (in a contributor field, you can add qualifiers to specify the exact role of the contributor), * to specify a quantity, a weight or a confidence level. This also allows the support of indexation chains like LCSH but keeping separate the individual concepts of the chain (see OCLC FAST project). You can then facetize on separate concepts and not whole chains. Current DSpace Authority List implementation is very different in data representation and corresponding abstraction: it puts the reference on the side in a separate database column together with a confidence level. Simpler/faster to implement but, IMHO, not adequate to provide the flexibility in indexing needed by Subject Repositories. Indexing can be more than a list of keywords: it can be "sentences" of coordinated concepts, "sentences" where Lucene Word Search engine brings nice searching functionalities (parameterized proximity search for instance). On my side, if a real interest comes from the Community and a knowledgeable committer like you want to help, I may: 1) gather the support of interested organizations 2) adapt the current code to the DSpace trunk 3) validate the result on real applications 4) foster a DSpace Subject Repository sub-community to build support around existing vocabularies and advance on real implementation of faceted indexing/browsing where users are put in contribution. Have a nice week! Christophe Dupriez DESTIN inc. SSEB [email protected] ------------------------------------------------------------------------------ This SF.net email is sponsored by Make an app they can't live without Enter the BlackBerry Developer Challenge http://p.sf.net/sfu/RIM-dev2dev _______________________________________________ Dspace-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/dspace-general
